cancel
Showing results for 
Search instead for 
Did you mean: 
688 Views
85 REPLIES 85

Shamino wrote:
actual received or vid?
you can trim downwards for vf8 as u know, if u have margin


Update : Ran a CB20 singel thread again, and 1.58 VID and 1.563v received as reported by AIDA64. Using +1 preset for OCTVB btw and not custom.
I guess this is to be expected from a SP63 chip.

Two BIOS screens to show V/F table and OCTVB:

dbavatar
Level 7
Shamino wrote:
you can tell the bios for the model from the link names:

https://www.dropbox.com/s/kl65p17gjpbiyx4/ROG-STRIX-Z490-I-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/xng2l0aupm72osv/ROG-STRIX-Z490-H-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/4esd4hdgnbmr0ra/ROG-STRIX-Z490-G-GAMING-WIFI-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/8q3632jjomgcz5m/ROG-STRIX-Z490-G-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/hlgyk8vsgc9y019/ROG-STRIX-Z490-F-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/xk9wssq60pep6xw/ROG-STRIX-Z490-E-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/cb2eq4npkkzeulm/ROG-STRIX-Z490-A-GAMING-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/b9xaltbtyxzkrb7/ROG-MAXIMUS-XII-HERO-WIFI-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/bux64zsst87ostu/ROG-MAXIMUS-XII-EXTREME-GUNDAM-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/bnj0o6o49xfn6al/ROG-MAXIMUS-XII-EXTREME-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/388yh4g9s0p6xz3/ROG-MAXIMUS-XII-APEX-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/a60fkicxurkc615/PRIME-Z490-A-ASUS-0901.rar?dl=0

https://www.dropbox.com/s/6f35mwxrv5knprv/ROG-MAXIMUS-XII-FORMULA-ASUS-0901.rar?dl=0



I'd like to point out that TVB was and is still broken on with this BIOS my i9-10850k as I previously detailed here: https://rog.asus.com/forum/showthread.php?121313-Z490-G-WIFI-multiple-BIOS-related-issues-memory-cpu...

Effectively, the two preselected "preferred" cores, in my case core 2 and 3, receive a 100mV VID boost (from ~1.360 to ~1.460), which cannot be defeated. I was hoping the new setting "cpu core auto voltage cap" would let me set a ceiling, but it as well is not effective. This has the result of quite hilariously running the preferred cores at a lower frequency, since the 100mV boost causes a temperature rise that throttles TVB frequencies! So now with 52x3 51x6 50x10 settings, all cores EXCEPT 2 and 3 boost to 52. Furthermore, this latest bios seems to break V/F control. My chip cannot run 53 on a specific core at stock voltages. Increasing the V/F setting for 53 by any amount takes no effect on any core! I could probably run 53 if I could reliably set 1.4V+ for that V/F, and not have core 2/3 over-volt. However it will only run at 1.356V, and I cannot get it higher.

I understand why most people might not notice these single/few core problems, it's because Windows is so noisy it's hard to actually test a single or specific core with a stress test and see a stable VID request. You can try that by pinning all processes on the machine to specific cores with this script run in a Administrator Powershell:

$instances = Get-Process; foreach ($i in $instances) { $i.ProcessorAffinity=0x100}

For 10850k/10900k this is a bitmap with valid values from 0x1 to 0xfffff. The above command selects a single core for all processes to run on, in this case on one of my "preferred" TVB cores. I can switch to another non-preferred core, and compare the VID (get and set, which are the same at this low load).

Switch back to core 0:
$instances = Get-Process; foreach ($i in $instances) { $i.ProcessorAffinity=0x1}

Switch back to all cores:
$instances = Get-Process; foreach ($i in $instances) { $i.ProcessorAffinity=0xfffff}

Finally, I scripted this to walk through all 20 cores and combinations of them every couple seconds while a stress test is running to validate that each core is stable at the boost frequency/voltage. I can post that if anyone finds it useful.

Shamino
Moderator
Effectively, the two preselected "preferred" cores, in my case core 2 and 3, receive a 100mV VID boost (from ~1.360 to ~1.460), which cannot be defeated. I was hoping the new setting "cpu core auto voltage cap" would let me set a ceiling, but it as well is not effective.
you cannot pin the proc's VID behavior on the bios.
the cap only caps the setting, the resultant v are determined by a whole slew of other factors.
fire up turbovcore or xtu read the adative voltage setting, it will not exceed what u cap it at. thats just the setting like i said.
yet i do not see how the v cannot be defeated. do a negative offset to vf pt 8. if it still doesnt drop, it may be due to vf7 being bigger than what you are trying to do to vf8
in this case. do it to vf 7 as well
for sure the voltage will drop if you do it per point

PerpetualCycle
Level 13
It seems it would be simpler if the adaptive mode's additional turbo voltage (ATV) were independent of the VF curve - that is, VF8 stuck at the native turbo multiplier and only ATV affected the maximum multiplier (above the native turbo). Then you would want the voltage for maximum multiplier to be the maximum of Vf8 (native turbo) and ATV.

For instance, I have a VF curve that is stable at 51x all-core. When I do a OCTVB+1 @ 52x on my 10700k, it drags VF8 to 52x. Since the VF is monotonic increasing, this means 51x is going to run at a lower voltage now instead of VF8+offset since it is interpolated. My OC is unstable not because of the 52x voltage, but because 51x is now running at a lower voltage, So I have to diddle with VF7 and VF8 to get 51x stable again in a non-optimal way. If the VF8 point stuck to 51x, then I would set the voltage I want for 52x via ATV and not have to tangle with the VF curve (ever again)

You must see some down side to this suggestion?

BTW, thanks for posting the tool for manipulating OCTVB, it is very useful for tuning without having to do constant reboots (which I am doing right now). 🙂


EDIT: I think what I am suggesting is the way adaptive used to work before the VF curve was exposed for editing.

ROG Hero XIII | 10900k @5.2 GHz | g.skill 2x32GB 4200 CL18 | ROG Strix 2070S | EK Nucleus 360 Dark | 6TB SSD/nvme, 16TB external HDD | 2x 1440p | Vanatoo speakers with Klipsch sub | Fractal Meshify 2 case

Have any of you guys tried Auto vcore when using this OCTVB feature?

geneo wrote:
It seems it would be simpler if the adaptive mode's additional turbo voltage (ATV) were independent of the VF curve - that is, VF8 stuck at the native turbo multiplier and only ATV affected the maximum multiplier (above the native turbo). Then you would want the voltage for maximum multiplier to be the maximum of Vf8 (native turbo) and ATV.

For instance, I have a VF curve that is stable at 51x all-core. When I do a OCTVB+1 @ 52x on my 10700k, it drags VF8 to 52x. Since the VF is monotonic increasing, this means 51x is going to run at a lower voltage now instead of VF8+offset since it is interpolated. My OC is unstable not because of the 52x voltage, but because 51x is now running at a lower voltage, So I have to diddle with VF7 and VF8 to get 51x stable again in a non-optimal way. If the VF8 point stuck to 51x, then I would set the voltage I want for 52x via ATV and not have to tangle with the VF curve (ever again)

You must see some down side to this suggestion?

BTW, thanks for posting the tool for manipulating OCTVB, it is very useful for tuning without having to do constant reboots (which I am doing right now). 🙂


EDIT: I think what I am suggesting is the way adaptive used to work before the VF curve was exposed for editing.


Vf pt addition never changed how adaptive works. It seems like intel firmware engineers take precautionary measures to ensure what is deemed illogical doesnt take place, hence all these conditions yhat need to be fulfilled before taking effect, but it certainly narrows down your options at times

Shamino wrote:
Vf pt addition never changed how adaptive works. It seems like intel firmware engineers take precautionary measures to ensure what is deemed illogical doesnt take place, hence all these conditions yhat need to be fulfilled before taking effect, but it certainly narrows down your options at times


Any frequency above 53x use the adaptive! If it is higher than V/F... This is from intel project...
Now I understood!
is it Right?

HELLO
I give you screenshot I use the bios 901 with AI OPTIMISER

RobertoSampaio wrote:
Any frequency above 53x use the adaptive! If it is higher than V/F... This is from intel project...
Now I understood!
is it Right?


Adaptive plus vf8 if vf8 is not 0 offset